This invention relates to microchannel plate (MCP) electron multipliers. In particular, the invention relates to conductively cooled MCPs which can be continuously operated at relatively high power levels without thermal runaway.
A channel electron multiplier 10 (FIG. 1) of the prior art is a device which detects and amplifies electromagnetic radiation. A secondary electron emitting semiconductor layer 12, which gives up one or more secondary electrons 14 in response to bombardment by primary radiation 16, for example, photons, electrons, ions or neutral species, is formed on the inner surface of the glass channel wall 18 during manufacture. Thin film metal electrodes 20 are deposited on opposite ends of the channel 18. A bias voltage 22 is imposed across the channel 18 to accelerate the secondary electrons 14 which are created by the incident radiation 16 at the input end of the channel. These electrons are accelerated along the channel until they strike the wall again, creating more secondary electrons. The avalanching process continues down the channel, producing a large cascade of output electrons 24 at the channel output.
A microchannel plate or MCP 30 (FIG. 2) of the prior art is an electron multiplier array of microscopic channel electron multipliers. The MCP likewise directly detects and amplifies electromagnetic radiation and charged particles. Currently a typical MCP is manufactured from a glass wafer 32 having a honeycomb structure of millions of identical microscopic channels 34, with a channel diameter which can be as small as a few microns. Each channel is essentially independent of adjacent channels, and is capable of functioning as a single channel electron multiplier. The channels 34 are coated with a semiconductor material 36. Active or respective input and output faces 38 and 40 of the MCP 32 are formed by corresponding apertured bias electrodes 42 and 44 which may be deposited by vapor deposition or sputtering techniques onto the wafer 32. The anode collector 50 is secured in confronting spaced relationship with respect to the output face 40 of the MCP 30 for collecting the electron output charge cloud or output 52. Typically, mounting apparatus 56 secures the microchannel plate 32 and the anode 50 in a vacuum chamber 54, and provides electrical connections 56 to the bias electrodes 42 and 44. After leaving the channel 34, the amplified charge cloud 52 is collected by one or more metal anodes 50 to produce an electrical output signal, or else impinges on a phosphor screen (not shown) to produce a visible image. By appropriate biasing of the electrodes 42 and 44 and the anode 50 the charged particles are driven from the MCP output to the anode across gap 62.
In general, the anodes or the phosphor screen are always separated from the output face 40 of the MCP 30. More sophisticated electrical readout configurations than simple anode pads include multi-wire readouts, multi-anode microchannel array (MAMA) coincidence readouts, CODACON, wedge and strip, delay line, or the resistive anode encoder. Although a direct contact anode has been mentioned in the literature, most conventional devices, including the aforementioned arrangements, require physical separation (i.e., gap 62) of the anode from the MCP output face.
Thermal radiation 60 emanating from the input face 38 as well as the output face 40 of the MCP 30 is the predominant and primary mechanism for transport of heat from the device 30. A small portion of the MCP heat 60' is conducted laterally through the MCP 30 to the metal mounting apparatus 56. According to the prior art, typical maximum heat dissipation of an arrangement such as is illustrated in FIG. 2 is limited to about 0.1 watt/cm.sup.2 of MCP active area as further discussed below.
As a sizeable electron cascade develops towards the end of the channel, secondary electrons lost from the channel wall leave behind a positive wall charge, which must be neutralized before another electron cascade can be generated. This is accomplished by the bias current flowing down the channel from the bias voltage supply (not shown), which also establishes the axial channel electric field. Neutralization must occur at a rate faster than the input event rate if multiplier efficiency is to be maintained, or else the multiplier gain will rapidly deteriorate and subsequent input events will not be sufficiently amplified. In effect, the channel is paralyzed, resulting in a channel dead time, the time required to neutralize the positive wall charge before the gain process can be reestablished.
Increasing the MC bias current decreases the channel dead time, hence it is desirable that the resistivity of the channel wall material be as low as possible while still maintaining its role as a potential divider. However, the semiconducting material on the channel wall exhibits a negative temperature coefficient of resistance (i.e, as temperature increases, resistance decreases.) Resistive (or joule) heating is caused by the flow of bias current. If this is not dissipated quickly enough from the MCP active area, it will lower the MCP resistance, resulting in increased bias current, which in turn will result in additional joule heating. (Use of voltage- or current-controlled power supplies cannot prevent this without changes to MCP gain.) Therefore if the initial MCP resistance is too low, thermal equilibrium will never be reached at operating voltages, and a critical temperature will soon be exceeded so that thermal runaway occurs and the MCP is destroyed.
In conventional MCP mounting configurations (FIG. 2) where the active areas of both MCP faces 40 and 42 are open to the vacuum, practically all the joule heat must be dissipated radiatively from the faces, since there can only be negligible conduction through the rim 63 to the mounting apparatus 56 due to the low thermal conductivity of glass. This inefficient heat removal process prevents thermal equilibrium from being reached at power levels greater than roughly 0.1 watt/cm.sup.2, which can be shown using the Stefan-Boltzmann law and appropriate values for MCP thermal emissivity. This corresponds to a maximum MCP bias current of about 100 microamps/cm.sup.2 at 1000 V, or a single channel resistance of roughly 10.sup.12 ohms.
This upper limit to MCP bias current will place a limit on the channel recharge time, limiting the MCP count rate capability or frequency response and thus dynamic range. For an output electron cascade of at least several times 10.sup.5 electrons, required for pulse-counting, the channel recharge time will be at least several milliseconds. If the count rate per channel exceeds about 100 Hz, the channel will be unable to recharge sufficiently, with a consequent degradation in gain and loss of multiplier efficiency. Assuming a channel packing density on the order of 10.sup.6 /cm.sup.2 and Poisson counting statistics, this places an upper limit to the overall MCP output count rate capability of roughly 10.sup.8 cts/cm.sup.2 /sec.
For an increasing number of applications, it is desirable to maintain pulse-counting gain beyond this upper limit, well into the gigahertz frequency region. This can only be achieved by increasing the bias current to a level where channel recharge times are on the order of several microseconds. However, this is obviously impossible using current MCP mounting configurations, where the primary means of heat removal must be through radiation.
In some applications a photocathode (not shown) is closely spaced in front of the MCP 30 to convert incoming visible and UV radiation into photoelectrons, which then act as the primary source of input radiation to the MCP. Photocathodes are quite heat sensitive and produce electrons spontaneously by thermionic emission. As the temperature of the MCP increases, the radiated heat is absorbed by the photocathode causing increasing amounts of spurious electron emission which are then amplified by the MCP, thereby resulting in noise at the output. This heat induced detector noise is undesirable.